You are on page 1of 22

PUBLICATIONS

Journal of Geophysical Research: Oceans


RESEARCH ARTICLE Unsteady stress partitioning and momentum transfer in the
10.1002/2014JC010240
wave bottom boundary layer over movable rippled beds
Key Points: S. Rodrıguez-Abudo1,2 and D. L. Foster3,4
 Double-averaging technique
1
implemented on PIV observations of Department of General Engineering, University of Puerto Rico-Mayaguez, Mayaguez, Puerto Rico, USA, 2Center for
the WBBL Applied Ocean Sciences and Engineering, University of Puerto Rico-Mayaguez, Mayaguez, Puerto Rico, USA, 3Department
 Partitioning of stress due to wave-
current, turbulence, viscosity, and
of Mechanical Engineering, University of New Hampshire, Durham, New Hampshire, USA, 4Center for Ocean Engineering,
bed forms University of New Hampshire, Durham, New Hampshire, USA
 Momentum balance in WBBL
provides with estimates of bed form-
induced pressure Abstract Observations of the nearbed velocity field over a rippled sediment bed under asymmetric
wave forcing conditions were collected using a submersible particle image velocimetry (PIV) system. To
Correspondence to: examine the role of bed form-induced dynamics in the total momentum transfer, a double-averaging tech-
S. Rodriguez-Abudo,
nique was implemented on the two-dimensional time-dependent velocity field by means of the full
rodriguez.abudo@upr.edu
momentum equation. This approach allows for direct determination of the bed form-induced stresses, i.e.,
stresses that arise due to the presence of bed forms, which are zero in flat bed conditions. This analysis sug-
Citation:
Rodrıguez-Abudo, S., and D. L. Foster gests that bed form-induced stresses are closely related to the presence of coherent motions and may be
(2014), Unsteady stress partitioning partitioned from the turbulent stresses. Inferences of stress provided by a bed load transport model suggest
and momentum transfer in the wave
that total momentum transfer obtained from the double-averaging technique is capable of reproducing
bottom boundary layer over movable
rippled beds, J. Geophys. Res. Oceans, bed form mobilization. Comparisons between the total momentum transfer and stress estimates obtained
119, 8530–8551, doi:10.1002/ from local velocity profiles show significant variability across the ripple and suggest that an array of sensors
2014JC010240.
is necessary to reproduce bed form evolution. The imbalance of momentum obtained by resolving the dif-
ferent terms constituting the near-bed momentum balance (i.e., acceleration deficit, stress gradient, and
Received 13 JUN 2014
bed form-induced skin friction) provides an estimate of the bed form-induced pressure that is consistent
Accepted 3 NOV 2014
Accepted article online 8 NOV 2014
with flow separation. This analysis reveals three regions in the flow: the free-stream, where all terms are rela-
Published online 13 DEC 2014 tively balanced; the near-bed, where momentum imbalance is significant during flow weakening; and below
ripple crests, where bed form-induced pressure is the leading order mechanism.

1. Introduction
Nearbed dynamics play a crucial role in the transport of sediments, transformation of waves across the shelf,
and beach evolution. The complex interaction between the seabed roughness and fluid forces imposed by
the waves and currents takes place in a relatively small portion of the flow domain (the bottom boundary
layer), which for free-surface gravity waves is confined to the lower few centimeters of the water column
[Grant and Madsen, 1986].
The dynamics within the wave bottom boundary layer (WBBL) are dictated by the resistive mechanisms
imposed by the various roughness scales that compose the seabed boundary. Over a flat sediment bed, the
roughness scales are given solely by the median grain diameter (d50), and the total stress can be partitioned
into contributions given by the viscous stresses acting on the sediment grains, the pressure drag imparted
by each individual grain, and the momentum transfer associated with turbulence and interactions between
the fluid and the sediment particles, collectively referred to as skin friction within the sediment transport
community [Engelund and Hansen, 1967; Nielsen, 1992]. Commonly, skin friction is estimated empirically and
semiempirically by either parameterizing the tractive forces with the wave friction factor [e.g., Maier and
Hay, 2009]; fitting a logarithmic model to velocity observations to get shear velocities [e.g., Fredsøe et al.,
1999]; or, mostly for modeling efforts, by characterizing the turbulence with a closure scheme, such as the
eddy viscosity [e.g., Grant and Madsen, 1979].
In the presence of roughness elements, with length scales much larger than the sediment grain diameter
(i.e., ripples or other types of bed forms), an additional stress mechanism emerges, generally called form
drag (or form-induced stress), resulting from the pressure gradients imposed by the rippled boundary. For
nonseparated sinusoidally varying flows and symmetric bed forms, these bed form-induced pressure

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8530
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

gradients would be expected to cancel when integrated over the ripple length. However, when these sim-
plified conditions are not met, there will be a net stress applied to the bed. In the presence of coherent
motions due to flow separation at the lee side of the ripples [Nichols and Foster, 2007; Van der Werf et al.,
2007], the local stress will exhibit a high variability across the ripple and can potentially complicate the
interpretation of observations resulting from individual velocity profiles. While some stress parameteriza-
tions exist to account for the geometry of the ripples (see e.g., hr in Nielsen [1992]), a comprehensive assess-
ment of the role of coherent structures in the total time-varying stress estimates, as given by a momentum
balance, provides an opportunity to further our understanding of the WBBL.
Temporal averaging, or Reynolds averaging, is a common technique for partitioning the independent con-
tributions of steady, periodic, and turbulent contributions to the total momentum flux [Reynolds and Hus-
sain, 1972]. An analogous decomposition in the spatial domain allows for an assessment of the impact of
uneven boundaries (see earlier geophysical efforts by Smith and McLean [1977] and Wilson and Shaw
[1977]). More recently, Gimenez-Curto and Corniero Lera [1996] (GC96) and others [Nikora et al., 2001; Hsu
et al., 2002; Maddux et al., 2003; Lowe et al., 2005] have incorporated all the dynamic effects of the seabed
roughness into the total stress field by double averaging the equations of motion (in both, time and space).
Thereafter, several authors have modified the double averaging technique to study different flow environ-
ments. Implementation of this technique on laboratory data requires at least two-dimensional observations
of the velocity field; therefore, its usage has been limited to relatively few efforts [e.g., Coleman et al., 2008].
In this study, the efforts of Coleman et al. [2008] are extended to yield a momentum balance between the
acceleration deficit, form-induced drag and stress gradient, as a function of wave phase. The continuity and
momentum equations describing the flow in the WBBL over a rippled sediment bed are averaged in time
and space resulting in the Double-Averaged Navier-Stokes (DANS) equations. This method allows for a
space-averaged stress partitioning scheme that considers the shear contribution due to viscosity plus the
phase-dependent momentum contributions due to waves and currents, turbulence, and the presence of
bed forms. In a manner similar to the so-called Reynolds stress, the term stress will be used throughout this
paper in reference to both, the viscous shear stresses and the actual momentum fluxes.

2. Theoretical Development
Following GC96, velocity and pressure are decomposed into phase-dependent and fluctuating components
defined by
~ i 1 u0i
ui  u (1a)

~ 1 p0
pp (1b)

where over tilde (~) denotes phase average and prime (0 ) represents departure from the latter. Phase-
averages (~t ) are calculated with

1X M
~ i ð~t Þ 
u ui ðt1mTÞ; (2)
M m50

where M is the total number of waves and T is the wave period. Please note that this term includes any
mean flow quantities. Throughout the rest of the paper the term phase will be used in reference to the tem-
poral variation of the wave.
Substituting (1) in the Navier-Stokes equations for incompressible flow in Cartesian tensor notation and tak-
ing the phase average yields

@~ ~ j Þ @ðug
ui u
u i @ð~ 0 u0 Þ
i j ~
1 @p @2u~i
1 1 52 1m ; (3)
~
@t @xj @xj q @xi @xj @xj

where tildes have been omitted from terms that are already a function of phase (~t ) and fluctuating quanti-
ties are assumed zero when phase-averaged over several wave periods. The equations presented in (3) are
the Reynolds-Averaged Navier-Stokes (RANS) equations for flows with a periodic signal [Reynolds and Hus-
sain, 1972].

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8531
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Following GC96, a second averaging operation for


flow over bed forms is introduced
ðð
1
Lo hFi 
Af
F dx1 dx2 ; (4)
Af

where F represents any spatially continuous flow


quantity, and angle brackets denote a spatial aver-
age over a horizontal plane with area Af defined in
the x1 (cross-shore) and x2 (alongshore) directions.
For bed forms uniform in the alongshore direction,
A B (4) reduces to
ð
1
hFi  F dx1 ; (5)
Lf
x3 Lf

x1 Lf (x3)
where Lf ðx3 Þ is the length occupied by the fluid in a
x1 = x'1a x1 = x'1b very thin slab of length L0, as shown in Figure 1. For
the simplified case of one single bed form, Lf
Figure 1. Schematic of the spatial averaging domain. For each represents the horizontal distance between points A
vertical location, 33, the average operation is performed over ðx 01a ; x3 Þ and Bðx 01b ; x3 Þ, such that Lf ðx3 Þ5x 01b 2x 01a .
the portion of the slab occupied by the fluid, Lf. The maximum
Above the ripple crests, flow quantities are assumed
length of the slab is L0, which in this case is equal to the dis-
tance between consecutive ripple crests, kb. Points Aðx 01a ; x3 Þ to be periodic in space over subsequent ripples, and
and Bðx 01b ; x3 Þ are located at the bed and are different at each Lf 5L0 5nkb , where n is the number of bed forms
vertical position. Above ripple crests, Lf ðx3 Þ5L0 .
included in the average, and kb is the bed form
wavelength. Please note that this formulation is only
valid if the variability of the wave forcing occurs over a significantly larger scale than the averaging scale,
i.e., k  Lf , where k is the length of the waves.
The spatial averaging operation given in (5) is applied to each term in (3). For cross-shore derivatives, the
spatial average becomes h@F=@x1 i5@hFi=@x1 50. While F can be any flow quantity, other than pressure
which will be considered later, for the case of velocity products in (5) h@F=@x1 i50. In the case of the along-
shore derivatives, uniformity in this direction also yields h@F=@x2 i50.
Spatial average over the vertical and temporal derivatives is slightly more complicated, as the limits of inte-
gration are functions of the differentiating variable. For a fixed bed and no-slip, no-percolation boundary
conditions, extraction of the partial derivative from the integrand is possible, resulting in the vertical deriva-
tive to be expressed as
 
@F 1 @Lf hFi
5 ; (6)
@x3 Lf @x3

and the temporal derivative to be


 
ui
@~ 1 @Lf h~
ui i
5 : (7)
@~t Lf @~t

However if we relaxed the boundary conditions and allow for a movable bed with migrating ripples, the
vertical derivative becomes
   
@F 1 @Lf hFi F dx 01
5 2 ; (8)
@x3 Lf @x3 Lf dx3 Lf

the temporal derivative results in


   
ui
@~ 1 @Lf h~
ui i ~ i dx 01
u
5 2 ; (9)
@~t Lf @~t Lf d~t Lf

and the momentum equations take the form of

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8532
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

 
u i i @Lf h~
@Lf h~ ~ j i @Lf hug
ui u 0 u0 i
i j Lf ~
@p @ 2 Lf h~
uii
1 1 52 1m 1Q; (10)
@~t @x3 @x3 q @xi @x3 @x3

where Q includes the second terms on the RHS of (8) and (9) times Lf.
In a manner analogous to a Reynolds decomposition, the mean and phase-averaged velocity and pressure
can be decomposed into a spatial average plus a bed form-induced disturbance, such that
~ i  h~
u u i i 1 ð~
u i Þb ; (11a)

~  hp
p ~ i 1 ðp
~ Þb ; (11b)

where angle brackets denote the spatial average given by (5), and the subscript b denotes contributions
attributed to the bed form disturbances. The inclusion of parentheses in the bed form component empha-
sizes the order of operation, where temporal averages precede spatial averages. While these averaging
operations are commutative, we have chosen time averaging before space averaging as it allows for a more
robust estimate of the bed elevation (see section 3). Please note that spatial decomposition of the turbulent
terms is not performed. The above definitions require the boundary disturbances to spatially average to
zero, i.e., hð~ ~ Þb i50.
u i Þb i5hðp
Substituting (11a) into the second term of (10) yields the spatial decomposition of the phase-averaged
velocity product
h~ ~ji
uiu u i i 1 ð~
5h½h~ u i Þb ½h~
u j i 1 ð~
u j Þb i
(12)
u i ih~
5h~ u j i 1 hð~
u i Þb ð~
u j Þb i;

where the definition hð~


u i Þb i5hð~
u j Þb i50 has been used.
To evaluate the pressure contribution, we consider linear free-surface gravity waves, where the variability of
~ occurs over a significantly larger scale than Lf, as k  Lf . The gradient @hp
p ~ i=@x1 is therefore nonzero and
can be approximated with the inviscid solution of the momentum equation in the free-stream, given, in the
cross-shore direction, by
u1i
@h~ ~i
1 @hp
2 ; (13)
~
@t q @x1

where velocity and pressure have been decomposed as in (11), with zero boundary disturbance, and the
pressure gradient assumed vertically uniform. While not included in the derivation, this formulation does
not preclude higher order wave solutions from being considered. To first order, the pressure contribution
given by the first term on the RHS of (10) becomes
 
Lf @ p~ @h~ ~ Þb j L f
u 1 i ðp
2 5Lf 2 : (14)
q @x1 @~t q

Substituting (12) and (14) into (10) yields the unsteady Double-Averaged Navier-Stokes (DANS) equations
for combined wave and current flow over a rippled bed, with horizontal component given in Cartesian coor-
dinates by

@Lf ðh~
u i2h~u 1 iÞ @Lf h~
u ihwi
~ @Lf hð~ ~ b i @Lf hug
u Þb ðwÞ 0 w0 i
1 1 1
@~t @z @z @z
(15)
~ Þb jLf
ðp 2
@ Lf h~ui
52 1m 1Q1R;
q @z 2

where x is the along-tank direction (1 offshore), y is the across-tank direction, z is the vertical direction, Q is
now given in Cartesian coordinates, and R52h~ u 1 i@Lf =@~t results from rearranging the first term on the RHS
of (14). Note that (15) assumes that the boundary irregularities are uniform in the alongshore direction and
allows for the presence of oblique currents (embedded in u ~ and w)
~ as long as they are alongshore uniform.
It may be rearranged such that
@Lf ðh~
u i2h~u 1 iÞ @L s
q ~ Þb jLf 1 f xz 1fsk 1f1 1fm ;
52ðp (16)
@~t @z

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8533
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

where sxz is the total stress given by


Table 1. Summary of Experimental Parameters
Parameter Value ui
@h~
sxz ðz; ~t Þ5 l 2qh~ ~ ~ b i2qhug
u Þb ðwÞ
u ihwi2qhð~ 0 w 0 i;
Wave height, H 5 cm @z
Wave period, T 2s (17)
Wavelength, k 3.3 m
Horizontal velocity amplitude, u0 11.8 cm s21 fsk is the bed form-induced component of skin fric-
Wave orbital semiexcursion, 3.8 cm
tion given by
A0 5u0 T=2p
Water depth, h 31 cm @ @Lf
Kinematic viscosity of water, m 1.004 3 1026 m2 s21 fsk ðz; ~t Þ5l h~
ui ; (18)
Sediment median grain diameter, d50 0.54 mm @z @z
Sediment specific gravity, s5qp =q 1.2
Sediment settling velocity, w0 1.8 cm s21 f1 relates to the temporal dependence of the
Ripple height, gb 0.8–1 cm averaging domain and is given by
Ripple wavelength, kb 4.5–7 cm
Migration rate, cb 20.005 cm s21 @Lf
Reynolds number, Re5u20 T=2pm 4400
f1 ðz; ~t Þ5qh~
u1i ; (19)
@~t
Particle Reynolds number, Res 5w0 d50 =m 9.7
2
Stokes number, Stk5pd50 s=9mT 0.06 and fm allows for ripples migrating and changing
Wave friction factor, 0.039
f2:5 5exp½5:213ð2:5d50 =A0 Þ0:194 25:977
shape such that
Grain roughness Shields 0.26   0  
parameter, h2:5 5f2:5 u20 =2ðs21Þgd50
dx d2 x 0 dx 0
fm ðz; ~t Þ5q u
~ 2m 2 2h~ ~
u wi : (20)
d~t dz dz Lf

This formulation also allows for the presence of suspended sediment particles as long as they follow the
streamlines closely with no significant interference with the flow.
The last two terms in (16) vanish for a fixed bed with no-slip, no percolation boundary condition, and the
momentum equation represents the balance between the acceleration deficit and the stress gradient plus
the form and viscous drag per unit fluid volume. The time-varying stress terms in (17) represent the viscous
stresses (first term), and the momentum transfer mechanisms due to the presence of free-surface gravity
waves (second term), uneven bottom boundary (third term), and turbulence (fourth term).
The expressions presented in (16)–(20) represent the x-component of the momentum equation in the
unsteady DANS formulation for WBBLs over a movable rippled bed, for the case of waves and ripples ori-
ented parallel to the shore. These expressions: (1) provide a momentum balance relating the acceleration
deficit to the force exerted by the ripples and sediment particles on the fluid, and the momentum transfer
gradients; (2) partition the stress allowing for the assessment of the contribution of individual terms; and (3)
allow for a sensible reduction of the two-dimensional time-varying observations. The following section
presents the results of an experimental investigation using this technique to better understand the mecha-
nisms associated with momentum transfer and bed form-induced pressure in the WBBL.

3. Observations
Detailed observations of the velocity field over a movable rippled bed were collected in a small-scale (1:15)
wave flume at the Fluid Mechanics Laboratory at Delft University of Technology, Netherlands, as described
in Rodriguez-Abudo et al. [2013]. The flume is 42 m long, 0.8 m wide, and 1 m high. Loose artificial sediment
grains, with specific gravity of 1.2 and mean grain diameter of 0.54 mm [Henriquez et al., 2008], were sub-
jected to regular waves, 5 cm in height and 2 s in wave period. The coordinate system of this study includes
x-positive offshore, z-positive upward and y-positive out-of-the-paper. Additional experimental characteris-
tics are outlined in Table 1 and Rodriguez-Abudo et al. [2013].
A Dantec Particle Image Velocimetry (PIV) system was used to resolve the nearbed 2-D velocity field and rip-
ple profile. Consistent with Nichols and Foster [2007], velocity vectors outside the boundary layer were vali-
dated with an Acoustic Doppler Velocimeter (ADV), time-synchronized with the PIV system, which consisted
of a 1 megapixel camera located outside the wave flume, and a 120 mJ ND:YAG laser located 27 cm above
the bed and roughly along the traversing center of the flume (please note that the expected boundary
layers at the walls are 1 cm thick) [Rodriguez-Abudo et al., 2013]. Image pairs were obtained over a 11 cm
3 11 cm sampling window at 12 Hz, with 10 ms between image pair members. Four 60 s data bursts were

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8534
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

collected, leading to a total of 720 velocity fields (30 waves) per burst. Seeding material included sediment
particles, micro-bubbles, and organic matter.
Velocity vectors were calculated with two passes and interrogation windows of size 64 pixels in the horizon-
tal and 32 pixels in the vertical, with 50% overlap, using the MatPIV 1.6.1 algorithms, developed by Sveen
[2004]. The resulting spatial resolution of the velocity field was 3.48 mm in the horizontal and 1.74 mm in
the vertical. Particle displacements were calculated at the subpixel level using a three-point Gaussian esti-
mator [Raffel et al., 2007]. Unrealistic velocity estimates (3.5% of the observations) were removed with a
three standard deviation filter and replaced with the local phase average, as described in Rodriguez-Abudo
et al. [2013]. Approximate uncertainty levels (60.72 cm s21, 66%u0 ) were computed with a bootstrap analy-
sis at the 95% significance level between the ADV and free-stream PIV signals [Nichols and Foster, 2007].
Please note that this estimate includes the error of both, the PIV and ADV velocities. Nearbed uncertainty is
approximated as 2.5%,pbasedffiffiffiffi on the particle number and trace size [Rodriguez-Abudo et al., 2013] and
should decrease by  M for the Reynolds stress estimates [Nimmo Smith et al., 2005].
As described in Rodriguez-Abudo et al. [2013], phase-resolved bed elevation estimates (zb ðx; ~t Þ) were com-
puted from both, the phase-maximum and phase-averaged pixel images. The lowest-elevation point with
the maximum pixel intensity in the phase-maximum pixel image (zbase ðx; ~t Þ) provided a baseline for the esti-
mate. The spatially averaged difference between the local centroid of the phase-averaged pixel image
(zc ðx; ~t Þ) and zbase was then added to zbase. The final zb was found by applying a running average of 30 pixels
(3.2 mm) to this quantity.
Imperfect periodicity of the waves and bed forms, high bed porosity, and three-dimensionality represent
sources of error that could potentially influence the final results of implementing the double-averaging
technique on the present observations. Departure from periodicity in the free-surface gravity waves results
in a deviation that may be assumed as normally distributed with zero mean [Coleman et al., 2008]. As the
number of waves increases this error will approach zero. Assuming that some waves are shorter and
others are longer with no apparent correlation or trends, then we also expect the potential effect of this
error on hug0 w 0 i to vanish over 30 wave periods.

The space average was performed over one ripple length to avoid introducing unnecessary noise into the
analysis due to low laser light on the upper left corner of the image (and therefore lower confidence of the
PIV estimates, Figure 2). By doing this, we assume that bed forms in the vicinity of the domain are well rep-
resented in our averaging interval (ergodic in space). This is clearly an approximation, as no bed forms are
alike, and therefore inferences such as the bed form-induced stresses may vary. The first and second
moments of the PDF of zb suggest that using one ripple length as the averaging domain is very similar to
using the full image (4% difference in the mean and 12% in the standard deviation). The measurement of
skewness yields a zero value for the central ripple length, and 20.4 for the entire image, which is a result of
the slightly deeper center trough. However, assessment of the swirling motions and bed form-induced
velocities (Figure 6) show that the dynamics are for the most part spatially periodic.
The large diameter of the scaled sediment used in the experiment may produce a higher than normal bed
porosity [Henriquez et al., 2008]. If there is a substantial interfacial flow within the grains, the assumptions of no-
slip no-percolation boundary conditions are not met, and a shear layer including the lower bound of the fluid
domain and the granular domain is likely to develop. The dynamics in the granular layer follows Darcy’s law
with a time modulation provided by the wave pressure gradient [Gimenez-Curto and Corniero, 2002]. However,
if pressure forces become 10% of the immerse weight of the grains, bed dilation and subsequent plug flow
conditions may occur [Sleath, 1999; Hsu and Hanes, 2004; Foster et al., 2006]. The theoretical development pre-
sented in section 2 allows for a slipping boundary (a persistently thick mobile layer consistent with visual obser-
vations). The effect of this phenomenon on the momentum balance will be further discussed in section 4.6.
A mean free-stream horizontal velocity of roughly 10.4 cm s21 is small and may be attributed to recircula-
tion in the wave flume (Figure 5b). As explained by Nielsen [1992], zero mass transport in the shoreward
direction of a wave flume produces a mean steady offshore current, known as Stokes drift, with magnitude
given by u stokes 5gH2 ð8chÞ21 , where g is the gravitational acceleration, H is the wave height, c5kT 21 is the
wave celerity, k is the wavelength, T is the wave period, and h is the water depth. For this data set,
H 5 5 cm, k53:3 m, T 5 2 s, and h 5 0.31 cm, resulting in u  stokes 5 0.6 cm s21, and consistent with the
free-stream mean flow. However, at the crest level h u i can reach up to 0.8 cm s21 in the offshore direction,

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8535
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 2. (a) Free-stream phase-averaged horizontal velocity (black), vertical velocity (red), and acceleration (blue). (b–i) Negative phase-
averaged image intensity with phase-dependent velocity field (red) for eight different phases indicated in Figure 2a. Scale vectors in the
top left corner represent 10 cm s21. The phase-varying averaging domain is contained within the two vertical solid lines shown in each
panel. Wave propagation is directed to the left.

consistent with the streaming profiles of Van der Werf et al. [2007]. Below the ripple crest, this value
decreases monotonically to 20.16 cm s21 at the trough.
The mean vertical velocity increases toward the bed, reaching a maximum value of 20.9 cm s21 at z53:1
cm. This is half the settling velocity computed with the formula of Cheng [2009]. The corresponding wave-
based Stokes number was 0.06 [Rodriguez-Abudo et al., 2013], and therefore, particles were assumed to fol-
low the streamlines closely (Stk << 1). Additionally, since sediment accretion was not observed, we con-
clude that our vertical velocity observations were not significantly affected by settling bias.
The theoretical development presented in section 2 assumes alongshore uniformity, thereby its implemen-
tation on PIV observations relies on two-dimensionality of the velocity field and bed form geometry. It is

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8536
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

likely that bed forms would have some variation in the across-tank direction, which could induce out-of-
plane motions and slightly three-dimensional vortex dynamics. Using a three-dimensional mixture model,
Penko et al. [2013] showed that for the present experimental conditions vortices formed and dissipated non-
uniformly in the cross-flow direction due to random fluctuations, yet swirling strength comparisons were
generally favorable. Although turbulent intensities measured in this effort are fairly consistent with previous
studies (see section 4.1), we anticipate that the effects of three-dimensionality may influence the Reynolds
stress estimates, which accounts for 20% of the total RMS stress (section 4.4).

4. Results
4.1. Temporal Decomposition
Phase-averages of the flow field, bed elevation, and image intensity provide with a general description of
the flow behavior, bed evolution, and nearbed sediment motion (Figure 2). Phase-averaged velocity and
image intensity were obtained from (2), while phase-dependent bed elevation was estimated using the
methodology described in Rodriguez-Abudo et al. [2013] with corresponding phase-averaged PIV images.
The following provides a general description of the flow, sediment, and bed dynamics observed on the
selected ensembles shown in Figure 2a:

1. At ~t 50 , right before on-offshore flow reversal, a small flow circulation area is observed on the onshore
(left) side of the ripples (Figure 2b). Bed forms appear asymmetric in shape, with wide crests, short lee
(onshore) slopes, and long stoss (offshore) slopes.

2. At ~t 545 , free-stream flow accelerated offshore, and nearbed velocities are slightly higher than the free-
stream velocities, especially above ripple crests and offshore slopes (Figure 2c). This is not inconsistent

with the overall 15 boundary layer phase lead reported in Rodriguez-Abudo et al. [2013]. Bed geometry
evolves into steeper ripples characterized by picked crests and deep troughs.

3. At ~t 5105 , the free-stream horizontal velocity is at an offshore maximum. This is consistent over the
entire observational window, with the exception of the offshore ripple slopes near the bed, where veloc-
ities have started to decelerate due to an adverse pressure gradient imposed by the ripple (Figure 2d).
Bed forms are more symmetrical in shape.

4. At ~t 5150 , the free-stream flow starts decelerating, and signatures of a boundary layer phase lead are
present on the onshore slopes of the ripples (Figure 2e). A velocity overshoot is present at about one rip-
ple height above the offshore ripple slopes, suggesting jet flow consistent with the observations of Van
der Werf et al. [2007].

5. Right before off-onshore flow reversal, at ~t 5195 , the free-stream acceleration is close to a negative maxi-
mum (Figure 2a) and a strong vortex is observed above the center ripple trough (Figure 2f). The bed
shape has now evolved to be slightly asymmetric with shorter lee (offshore) slopes, and longer stoss
(onshore) slopes.

6. At ~t 5240 , the free-stream flow is accelerating, jet flow is observable above the leftmost ripple crest, and
boundary layers develop at the offshore slope of the ripples (Figure 2g). The bed elevation shows sym-
metric bed forms.

7. By ~t 5285 , the flow has initiated horizontal deceleration, with some of the nearbed flow leading the free-
stream (Figure 2h). The bed geometry starts turning asymmetric again.

8. By ~t 5330 , the free-stream horizontal acceleration is at a positive maximum (Figure 2i). Ripple asymmetry
increases. Nearbed flow seems to undergo an additional stage of acceleration.
 
Figure 2 shows that the vortex formed at ~t 5195 is stronger than the one present at ~t 50 . In the study by
Van der Werf et al. [2007] with pure horizontally oscillating flow (no free-stream vertical velocity), stronger

vortices were observed near the on-offshore flow reversal (~t 50 in this study). The discrepancy can likely be
attributed to slightly skewed and asymmetric wave forcing [Rodriguez-Abudo et al., 2013], and also that, in
the present study, ripple slopes (gb =kb ), coincident with on-offshore flow reversal, are more gentle than
their counterparts due to large downward velocities, thus creating a weaker vortex. Additionally, wave verti-
cal velocities are not present in studies with oscillatory water tunnels, like that of Van der Werf et al. [2007].
Nichols and Foster [2007] also observed stronger vortices near off-onshore flow reversal. Their full-scale

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8537
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Phase [°]
a b c
11 360

10 315

9 270

8 225

Z [cm] 7 180

6 135

5 90

4 45
5 mm
3 0
2 4 6 8 10 −2 0 2 0.95 1
X [cm] cb [cm s−1] ρ

Figure 3. (a) Bed elevation as a function of wave phase (right vertical axis). Each bed profile is vertically offset by 3 mm (see gray dashed
line). (b) Bed form migration rate between consecutive ensembles. (c) Cross-correlation coefficient for each migration estimate.

experiment included wave groups and asymmetric ripples. While some of the high suspended sediment
concentration areas (visualized by the darker areas in Figure 2) are consistent with the presence of coherent
 
motions (e.g., ~t 5195 ), some other instances of suspended sediment lag the free-stream flow (e.g., ~t 5195 ).
This may be due to suspensions remaining from previous events, as observed in the field effort by Dick
et al. [1994], where an ensemble average technique based on peak suspended sediment concentration was
employed.
The relatively light sediment particles resulted in a dynamic bed geometry (Figure 3a). Intrawave bed form
migration rates obtained with cross-correlation analysis of consecutive phase-dependent bed elevation (Fig-
ure 3b) show across shore modulation of the bed that is in phase with the free-stream velocity. Such a
dynamic bed response has been observed previously in the efforts of Nichols and Foster [2009] and Hanes
et al. [2001]. The net migration (0.005 cm s21 directed onshore) was estimated by integrating the migration
over the wave phase (~t ).
Sleath [1999] argued that over flat beds, bed dilation occurs when the overlying wave acceleration is large.
In these observations, the bed modulates within the wave cycle. The ripple reforms (Figure 2c) after it has
been flattened during high acceleration and relatively strong negative wave vertical velocities (Figures 2b
and 2i). This modulation does not occur during the highest acceleration near the second flow reversal (Fig-
ure 2f). Interestingly, near both flow reversals (Figures 2b and 2f), the bed forms become more asymmetri-
cal, similar to those observed in unidirectional flow [Raudkivi, 1966], with long, mild stoss slopes, and short,
steep lee slopes.
Consistent with previous studies [Hino et al., 1983; Sleath, 1987; van der Dominic et al., 2011], phase-
dependent estimates of u0rms ðx; z; ~t Þ and w 0rms ðx; z; ~t Þ are stronger for phases of high inertia, peaking during
onshore directed flow (Figure 4), and significantly diminishing near phases of flow reversal. During peak
 
flows (~t 5105 and 285 ), turbulent intensities are larger above the lee slopes, likely due to the presence of
an adverse pressure gradient, and consistent with the numerical simulations of Barr et al. [2004]. Horizontal
turbulent intensities seem to dominate during high-flow phases, as shown by the nearly horizontal vectors.
Of particular importance to the stress partitioning scheme given in (17) and discussed further in the paper,
is the behavior of the phase-dependent turbulent velocity covariance (ug 0 w 0 ), as it represents the turbulent
g
0 0
contribution to the total stress. In general u w follows the same spatial and temporal characteristics as the

turbulent intensities. However, for ~t 5195 , when a strong circulation area is evidenced in both, Figures 2f
g0 0
and 6f, u w is relatively small, suggesting that turbulence can be partitioned from the bed form-induced
motion.

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8538
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 4. (a) Free-stream phase-averaged horizontal velocity (black) and vertical velocity (red). (b–i) Phase-dependent turbulent intensity
(u0rms ðx; z; ~t Þ and w 0rms ðx; z; ~t Þ, vector field) superimposed over jug
0 w 0 j (cm2 s22, gray scale) for eight different phases indicated in Figure 4a.

Scale vectors in the top left corners represent 10 cm s21. Wave propagation is directed to the left.

4.2. Spatial Decomposition


Spatially and phase-averaged horizontal velocities (h~ u i, Figure 5) show the behavior of typical oscillating

boundary layers. A phase lead of 15 , computed with cross-correlation analysis, is present at the lower
0.7 cm of the water column, consistent with the numerical effort of Penko et al. [2013], and previous labora-
tory [van der Dominic et al., 2011], and field [Foster et al., 2000] observations over a flat bed. Figure 5a shows
that in general, h~
u i is O(10) times larger than hwi.
~
The bed form-induced flow pattern (Figure 6) starts with a significant signature during negligible free-
 
stream forcing (~t 50 ). At ~t 545 , the presence of bed forms manifests as a positive ð~
u Þb above the ripple
crests, and is consistent with flow acceleration due to favorable pressure gradients. Vertical velocities point
upward as flow moves up the ripple slope, and downward at the lee slopes. The pressure gradient imposed

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8539
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 5. Double-averaged (in phase and space) wave velocity profiles (h~u i i) as a function of wave phase. The dark and light gray shaded
areas represent elevations below the deepest trough level and highest crest level, respectively, for each wave phase. Scale vectors on the
21
top left corner represent 6 cm s .


by the ripples is more apparent at ~t 5105 , as shown by the positive and negative values of ð~ u Þb and ðwÞ~ b
for flow constricting up the ripples’ onshore slopes and expanding down the offshore slopes, respectively.

As the free-stream flow begins decelerating (~t 5150 ), a wake-like area characterized by positive ð~ u Þb and

negative ðwÞ ~ b forms offshore of the ripple crests. Right before off-onshore flow reversal (~t 5195 ), a vortex
signature is clear in the ð~
u Þb and ðwÞ
~ b signals, suggesting that its existence is almost solely due to the pres-

ence of bed forms. During onshore flow strengthening (~t 5240 ), a boundary layer develops at the offshore
slope of the leftmost ripple crest, and its effect on ð~u Þb is illustrated by the positive vectors. As flow weakens

(~t 5285 ), the pressure gradient imposed by the bed forms is evidenced by flow acceleration (jet flow)
above the rightmost ripple crest, and reversing ð~ u Þb after flow passes the ripple crests. By the second flow

reversal (~t 5330 ), a new circulation area is present at the center trough, as clearly shown by ð~ u Þb and ðwÞ~ b.
In summary, the spatial decomposition of the flow suggests that (1) the spatially averaged flow leads the

free-stream by 15 ; (2) the presence of bed forms significantly influence the dynamics within roughly one
ripple height above the crests, which is consistent with the region of amplification of spectral energy found
by Rodriguez-Abudo et al. [2013]; and (3) vortices with scales similar to the ripple height occur during phases
of flow reversal, and are manifested in the bed form-induced velocity signal.

4.3. Coherent Structures


Following Nichols and Foster [2007], the large scale (O(gb)) coherent structures were characterized with a
swirling strength criterion [Zhou et al., 1999]. The slightly longer offshore-directed half-wave period, charac-
terized by steeper ripples (section 4.1), produces a stronger signature of coherent motions (Figures 6d–6f),
as previously observed by Nichols and Foster [2007], and consistent with the bed form-induced flow field.

These vortical structures begin their formation by maximum flow (~t 5105 ), and progressively strengthen
 
until they are ejected at flow reversal (~t 5195 ). A new set of structures develops at ~t 5240 , but weakens
through the half-wave period, likely due to reconfiguration of the bed, as it flattens reducing bed form-

induced pressure gradients. During flow weakening (~t 5330 ), a mild signature of coherent structures

develop at the onshore slopes of the ripples. By flow reversal (~t 50 ) only the structure associated with the
steeper ripples is maintained. Consistent with observations by Nichols and Foster [2009], this assessment
shows that steeper bed forms cause stronger and bigger vortices. The location and timing of these vortical
structures have been qualitatively validated with the numerical effort of Penko et al. [2013].

4.4. Stress Partitioning


Time and space-decomposition of the velocity field allows for individual estimates of the momentum flux
terms responsible for transferring momentum into the sediment bed (see expression 17). These terms rep-
resent: (1) large-scale momentum transfer mechanisms due to waves and currents (2qh~ u ihwi);
~ (2) bed
form-induced momentum transfer mechanisms (2qhð~ u Þb ðwÞ
~ b i); (3) phase-varying turbulent (or Reynolds)
stresses (2qhug 0 w 0 i); and (4) viscous stresses (l@h~
u i=@z). Even though most of these terms denote

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8540
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 6. (a) Free-stream phase-averaged horizontal velocity (black) and vertical velocity (red). (b–i) Bed form-induced velocity field (red
vectors) superimposed over the swirling strength (s21, gray scale) for eight different phases indicated in Figure 6a. Scale vectors in the top
left corners represent 5 cm s21. Wave propagation is directed to the left.

momentum transfer mechanisms, in this effort we will also refer to them as stresses, owing to their dimen-
sional similarities.
A simple order of magnitude assessment shows that the stress due to viscosity is significantly lower than
most of the other terms (Figure 7). Wave-induced stresses, resulting from correlations between horizontal
and vertical wave motions (2qh~ u ihwi,
~ Figure 7c), are significant in the free-stream, except at the horizontal
and vertical flow reversals. These arise from the mean acceleration given by the wave orbital motions and
agree with the trends observed by Coleman et al. [2008]. Near the bed (z5425 cm), 2qh~ u ihwi
~ is more pro-
nounced during onshore-directed half-wave period. In the RMS (Figure 9), 2qh~ u ihwi
~ decreases downward
until it reaches z55 cm (gb above the ripple crest), with a local minimum of 0.24 N m22 at the ripple crest,
and a local maximum of 0.36 N m22 below the crest. This represents a departure from common wave

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8541
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 7. (a) Free-stream phase-averaged horizontal (black) and vertical (red) velocity. (b–e) Terms on the right-hand side of (17) contribut-
ing to the total stress sxz, as a function of vertical position and wave phase. The lower bound of the domain is located at the ripple trough,
and the solid line represents the vertical position of the highest crest.

u ihwi
theories, which predict a continuous decrease in 2qh~ u ihwi
~ toward the bed. 2qh~ ~ represents 55% of
the total RMS momentum transfer near the bed.
The bed form-induced stress is only present in the nearbed region (Figure 7d). It accounts for 22% of the
total stress at the ripple crest level (Figure 9). Consistent with the analysis of Coleman et al. [2008], the bed
form-induced stress is negligible above one ripple height from the crest.
In general, the phase-varying Reynolds stress is confined to the region below the ripple crests, except for phases
with strong free-stream velocities, when it peaks and slightly propagate upward (Figure 7e). After off-onshore
flow reversal, the portion of 2qhug 0 w 0 i located below ripple crests switches sign before the portion above. In the

RMS, the Reynolds stress behaves very similar to the bed form-induced stress, accounting for 22% of the stress
at the ripple stress level and increasing downward (Figure 9). These results are different from slightly deeper water
observations of Coleman et al. [2008], who found phase-varying values of Reynolds stress to be negligible at all
heights and wave phases. Once again, these results suggest that small-scale turbulent momentum transfer may
be partitioned from the ripple-scale momentum fluxes provided by the bed form-induced coherent motions.
In contrast with the development of GC96 and measurements of Coleman et al. [2008], these results suggest
that turbulent stresses are significant. In analogy to the Reynolds stress, the bed form-induced stresses owe

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8542
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 8. (a) Free-stream phase-averaged horizontal (black) and vertical (red) velocity. (b–i) Total stress as given by (17) (black), and local
stress as given by (21) for x 5 2.5 cm (blue), x 5 4 cm (red), and x 5 5.5 cm (gray). The horizontal dashed line represents the vertical loca-
tion of the highest crest.

their existence to correlations in the spatially disturbed flow field, and therefore the large-scale coherent
structures should only show up in the bed form-induced stress signal. Below the ripple crest, the spatially
averaged swirling strength (Figure 10b) is consistent with the magnitude of qhð~ u Þb ðwÞ
~ b i (Figure 10c). They
also generally agree above the crest level, except during phases of high offshore-directed flow. The discrep-
ancy cannot be attributed to the Reynolds stresses, as there are clearly no similarities between hkci i and 2q
hug0 w 0 i (Figure 7e).

It was hypothesized earlier in the paper that local stress estimates provided by a single velocity profile are
not capable of reproducing the total momentum transfer on a rippled bed. The spatial variability of sxz is
examined with the local stress estimates at a crest, slope, and trough of the ripples (Figure 8). Local stress
estimates are computed from the RANS equations for flows with a periodic signal:

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8543
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

d~
u
Txz ðx; z; ~t Þ5l 2q~ ~
u w2qug
0w0 ; (21)
dz
where over tildes and primes are as defined in
section 2.
As expected, for elevations above one to two gb
above the mean ripple crest, the momentum
transfer is uniform across the ripple length. This
is not the case for the lower portion of the
domain, where local stress estimates can differ
significantly, especially during phases of flow
 
weakening (~t 5150 and 285 ). The strongest dis-
Figure 9. Contributions to the total RMS stress given by the individ- crepancies are found for profiles near the ripple
ual terms in (17). The horizontal black line indicates the average verti- trough (x 5 5.5 cm). While phases near the
cal location of the highest crest.
onshore-offshore flow reversal do not show sig-
 
nificant discrepancies (~t 5330 and 0 ), stress

profiles separated by only 2.5 cm can have opposite directions during the second flow reversal (~t 5195 ).
  
Additionally, strong signals of Txz above the ripple crest at phases of high inertia (~t 5105 ; 240 , and 285 )
are consistent with the presence of coherent motions (Figure 10b).

4.5. Total Stress


Shear stress has been extensively linked to the initiation of motion and subsequent sediment transport. In
order to assess the veracity of the total stress estimate (sxz) inferred from the DANS formulation, we

Figure 10. (a) Free-stream phase-averaged horizontal (black) and vertical (red) velocity. (b) Space-averaged swirling strength, and (c) mag-
nitude of the bed form-induced stress as a function of vertical position and wave phase. The lower bound of the domain is located at the
rippled trough, and the black solid line represents the vertical position of the highest crest. (d) sxz;rip (1), and hsMPM i (o) as a function of
wave phase (left axis), with corresponding Shields parameter (right axis).

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8544
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 11. (a) Comparison between sxz;rip and hsMPM i. Dashed line represents a 1:1 relationship. (b, c) sxz;rip and hsMPM i plotted against
1
u 1 ij. Solid lines represent linear regression models. Confidence limits on fw are obtained at the 90% level. All axes are in N m22.
u 1 ijh~
2 qh~

compare the vertically averaged sxz computed between the ripple trough and crest and weighted by f5Lf =
L0 (sxz;rip ) to an independent estimate of stress obtained from a simple bed load transport model [Meyer-
Peter and Muller, 1948]:
 
Qðs21Þg 2=3
sMPM 5q ; (22)
AMPM

where AMPM is a constant with an assumed value of 10, and Q is the sediment transport rate given by

Qðx; ~t Þ5Q0 1ncb ð~t Þzb ðx; ~t Þ; (23)

where Q0 is the sediment transport at the trough level and assumed zero based on visual observations,
n 5 0.7 is the sediment concentration within the bed form, cb is the bed form celerity, and zb is the bed
elevation.
sxz;rip and the spatially averaged stress inferred from the bed load transport model (hsMPM i) show good
agreement in both magnitude and phase, especially during the second half-wave period (Figure 10d). For ~t

502100 ; sxz;rip is roughly half the stress required by the [Meyer-Peter and Muller, 1948] model. The agree-

ment at ~t 52002250 is particularly remarkable given the high horizontal acceleration present at this wave
stage (Figure 2a), which has been shown to significantly enhance sediment dynamics [Sleath, 1999]. While
there can be some concerns with this comparison, given that sxz;rip includes a mean acceleration term, our
results show a fairly good agreement in both magnitude and phase. In general, the mobility of the bed
forms suggests a stress estimate 14% larger than sxz;rip (Figure 11a).
The stress, or total momentum, applied to the sediment bed should be correlated to the local sediment
transport and bed state (e.g., ripples). Commonly, the impact on the sediment transport is coupled with the
nondimensional Shields parameter, defined as the ratio of the shear stress to the immerse weight of the
grains:
s
h5 ; (24)
ðs21Þqgd50

where s is the averaged nearbed stress given by sxz;rip , s is the specific gravity of the grains, and q is the den-
sity of water. Overall, the phase-dependent h suggests that the bed remains within the bed form regime for

the entire wave period (Figure 10d). The largest shear stress occurs at 315 , when h reaches a peak value of
roughly 0.38 in the onshore direction, which is less than the threshold for sheet flow (50:8). This peak mag-
nitude is roughly 46% larger than the grain roughness Shields parameter (h2:5 50:26) and 2–3 times larger
than estimates found with a quasi-steady analysis [Rodriguez-Abudo et al., 2013].
The wave friction factor can be computed by assuming a quadratic stress law ~s 5 f2w qh~ u 1 ijh~
u 1 ij, where
both h~u 1 i and ~s are allowed to vary in time [Jonsson, 1966]. Here we assume that h~ u 1 i is a reasonably
good approximation of the nearbed velocity right outside the boundary layer. The resulting friction factor is
obtained by linearly regressing both stress estimates (i.e., sxz;rip and hsMPM i) to 12 qh~
u 1 ijh~
u 1 ij (Figures 11b–
11c). The analysis using sxz;rip shows less scatter, higher skill, and narrower confidence interval for fw than

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8545
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Figure 12. (a) Free-stream phase-averaged horizontal (black) and vertical (red) velocity. (b–e) Terms constituting the momentum balance
presented in (16) as a function of vertical position and wave phase. The lower bound of the domain is located at the rippled trough, and
the solid line represents the vertical position of the highest crest.

using hsMPM i, even though both estimates of fw are fairly close in magnitude. The resulting fw using sxz;rip is
36% higher than the wave friction factor proposed by Swart [1974] (f2:5 5 0.039; Table 1), which is attributed
to scaling effects [Henriquez et al., 2008]. Previous experimental [Carstens et al., 1969; Lofquist, 1980] and the-
oretical (GC96’s expression 45) efforts suggest values for fw up to five times higher than in the present study
(after correcting for the respective friction factor definitions). The discrepancy is intriguing and requires fur-
ther investigation, yet an assessment of the bed state using (24) and GC96’s (20) and (45) yields conditions
well above the ripple regime (h  2), which are unrealistic for this effort.

4.6. Momentum Balance


The terms within the momentum balance presented in (16) can now be evaluated individually as a function
of their vertical position and wave phase (Figure 12). For simplicity, the terms fsk, f1 , and fm have been
grouped together, as their magnitude is small compared to the other terms. As expected, a negative accel-
eration deficit dominates within wave phases with positive acceleration (see Figure 2 for a time series of
acceleration). Similarly, a positive acceleration deficit exists for wave phases with negative acceleration. The

latter is stronger than its counterpart due to wave steepness. Approximately 50 before both flow reversals,
 
an area of slight acceleration deficit develops above the ripple crests (~t 5140 ; z5 4–5 cm and ~t 5300 ; z5
4–5 cm), and almost completely vanishes by the time of flow reversal. This coincides with the layer of

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8546
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

lagging velocities observed in Figure 5, and suggests that deceleration in this layer is smaller in magnitude
than in the free-stream for phases of flow weakening.
Positive stress gradients are present below ripple crests during offshore-directed flow, and are propagated
upward after flow reversal. Similarly, negative stress gradients are present below the crests during onshore-
directed flow and fade away after the second flow reversal, reappearing above the crests and peaking dur-
ing the weakening of the offshore-directed flow. Below the ripple crests the stress gradient is, for the most
part, different in magnitude and direction from the same quantity in the region 1–2 cm above the ripple
crests. Therefore the stress gradient reaches its minimum magnitude near the ripple crest elevation. This is
consistent with the conceptualization that different momentum transfer mechanisms exist below and
above the ripple crests.
The bed form-induced skin friction (fsk), resulting from rearrangement of the viscous term in (15) (not shown
individually in Figure 12), is almost two orders of magnitude lower than the other terms in the momentum
balance. As expected, this term appears only at heights below the ripple crest, suggesting a decoupling
between these two regions. Similarly, fm is about one order of magnitude lower than the other terms in the
momentum balance. This term appears as a result of migrating ripples and bed form modulation. Even
though, both of these processes are relatively significant for this data set (Figure 3), fm does not contribute
significantly to the momentum balance. Perhaps the most significant term in Figure 12e is f1 , which is a
result of a time-varying averaging domain. It is more prominent during phases of flow acceleration
   
(~t 540 2150 ) and high inertia (~t 5225 2325 ), although it is still low when compared to the other more
complex dynamics shown by the acceleration deficit, stress gradient, and pressure. In summary, the low
magnitude of fsk 1f1 1fm suggests, not surprisingly, that vertical gradients in the momentum transfer and
form-induced pressure differences are the leading order mechanisms responsible for the deficit in accelera-
tion in this environment.
The imbalance of momentum contains the form-induced pressure signal (2ðp ~ Þb jLf ), observational and
numerical errors, and any physics not otherwise accounted for. While favorable pressure gradients at the
stoss side of the ripples induce flow acceleration, adverse pressure gradients on the lee side result in flow
deceleration and the formation of a separation bubble, characterized by a significant drop in pressure on
the lee face and a pressure rise near the reattachment point [Haque and Mahmood, 1983; Shen et al., 1990;
Fernandez et al., 2006]. The momentum imbalance below the ripple crest is consistent with this model
~ Þb ðx5x 0a ; zÞ2ðp
~ Þb jLf 5ðp
(recall that 2ðp ~ Þb ðx5x 0b ; zÞ, where x5x 0a is at the lee slope of the ripple during
offshore-directed flow, and at the stoss slope during onshore-directed flow), and agrees with the direction
of the phase-dependent bed form migration (Figure 3). These results suggest that this momentum imbal-
ance may provide an estimate for the bed form-induced pressure.
By definition, 2ðp ~ Þb jLf must be zero above the crest level for symmetric ripples. The imbalance of momen-
tum above this elevation during phases of flow weakening, suggests the existence of an additional mecha-
nism not quantified in this effort that may be responsible for the consistent lag of spatially averaged
velocities at this elevation. Examination of Figure 10b suggests that this imbalance is not related to the
coherent structures. A possible explanation is the lack of a vertically uniform wave-induced pressure gradi-
ent, as assumed in (13). Excess momentum associated with the wave-induced radiation stress, not included
in our derivation, is also a possible mechanism, worth of study for future efforts.
Examination of the resolvable terms in (16) suggests the existence of three different regions in the flow. The
first is located at the free-stream and, as expected, shows negligible acceleration deficits and relatively low
stress gradients. The second region is located 1–1.5gb above the ripple crests. In this layer, an imbalance of
momentum is characteristic for phases of flow weakening. The third layer encompasses the flow between
the ripple elements and is markedly different from the other two. In this layer, the leading order mechanism
controlling the dynamics during high flow phases is the bed form-induced pressure.

5. Discussion
Throughout this study, the term u ~ i have encompassed both waves and currents. Although mean flows are
small in this effort (see section 3), it is perhaps appropriate to further decompose the periodic (waves) and
mean (currents) signals in a manner similar to Coleman et al. [2008]. Recognizing that this extra

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8547
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

decomposition inherently results in terms which spatial scales are not well separated, in this section we
present the effects on the main results of systematically partitioning waves and currents.
Following Nielsen [1992], velocity and pressure are decomposed into mean (current), phase (wave), and fluc-
tuating (turbulent) components defined by

ui  u u i 1u0i ;
 i 1^ (25a)

 1p
pp ^ 1p0 ; (25b)

where overbar (2 ) represents the mean, over hat (^) denotes phase average with zero mean, and prime (0 )
represents departure from the latter. Time (t) and phase (~t ) averages are calculated following Nielsen [1992]:

1X N
i 
u ui ðtn Þ; (26)
N n51

1X M
^ i ð~t Þ 
u ui ðt1mTÞ2
ui; (27)
M m50

where N is the total number of samples, M is the total number of waves, and T is the wave period. After
phase and spatially averaging the equations of motion (with (27) and (5), respectively) and decomposing
i; u
u  , and p
^i; p ^ in space (in accordance to (25)), the stress term presented in (17) becomes

ui
@h^
sxz ðz; ~t Þ5 l ud
2qh^ ihwi2qh^
^ u ihwi2qh
 u ihwi
^
@z (28)
u Þd
2qhð^ ðwÞ
^ i2qhð^
b bu Þ ðwÞ
 i2qhðb ^ i2qhud
u Þ ðwÞ
b
0 w 0 i;
b b

where the first term on the RHS denotes the viscous stress, terms 2–4 represent the convective stresses
(waves and currents), terms 5–7 represent the bed form-induced stresses, and the last term is the Reynolds
stress. The quantities fsk, f1 , and fm become
@ @Lf
fsk ðz; ~t Þ5l h^
ui ; (29)
@z @z
@Lf
f1 ðz; ~t Þ5qh^
u1i ; (30)
@~t
"   #
dx 0 d2 x0 dx 0
fm ðz; ~t Þ5q u
^ 2m 2 2ðh^ ^
u wi1h ^
u wi1h^ 
u wiÞ ; (31)
d~t dz dz
Lf

respectively. Please note that the terms inside the square brackets in (31) must be phase averaged following
(27). The momentum balance is then given by
@Lf ðh^
u i2h^u 1 iÞ @L s
q ^ Þb jLf 1 f xz 1fsk 1f1 1fm :
52ðp (32)
@~t @z
Stress terms and momentum transfer mechanisms are quite similar in temporal and vertical pattern (figures
for decomposition 25 not shown). However, they differ slightly in magnitude. In Table 2, the temporal RMS
of several mechanisms computed at the ripple crest elevation using decompositions (1) and (25) is pre-
sented, as well as resulting parameterizations such as the Shields parameter and the friction factor.
The stress mechanisms computed with decompositions (1) and (25) at the ripple crest elevation differ by 4–
7%, with the highest difference showing in the Reynolds stress (Table 2). For the case of the convective, bed
form-induced, and total stresses the slightly lower estimates may be attributed to the fundamental differ-
ence between definitions (2) and (27), where the latter represents a phase average with zero mean. How-
ever, for the case of the Reynolds stress, the phase-average definition does not quite explain the percent
difference. Since the Reynolds stress encompasses any phenomena remaining from the waves and currents
(noise included), it is perhaps safe to attribute the difference to experimental uncertainty, among others.
For the case of the mechanisms conforming the balance of momentum at the ripple crest, the difference
between using (1) and (25) lies within 10–26%. The highest discrepancies are found for the stress gradient.

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8548
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Table 2. Summary of Results Computed With Decompositions (1) and (25)


Decomposition (1) Decomposition (25)
a 22 21
Convective stress [N m ] 2.4 3 10 2.3 3 1021
Reynolds stressa [N m22] 9.4 3 1022 8.8 3 1022
Bed form stressa [N m22] 8.4 3 1022 7.9 3 1022
sxz a [N m22] 3.4 3 1021 3.2 3 1021
%Diff. (sxz;rip , sMPM) 14 20
hrms 0.23 0.33
hmax 0.38 0.41
fw 0.053 6 0.009 0.055 6 0.008
Acceleration deficita [N m22] 4.2 3.8
Bed form-induced pressurea [N m22] 4.3 3.6
Stress gradienta [N m22] 1.3 1.0
fsk 1f1 1fm a [N m22] 1.8 1.6
a
RMS computed at the ripple crest elevation.

Please note that at the ripple crest elevation the stress gradient is a minimum, as the stress generally
switches sign at this elevation. The percent difference in the stress gradient, unlike the other terms in the
momentum balance, is well explained by the difference between phase-average (2) and (27).
The mobility of the bed forms provide a stress estimate (sMPM) that is slightly larger than sxz;rip by 14% and
20% for decomposition schemes (1) and (25), respectively. Once again, this may be a result of including the
mean contributions in the phase average (2). For the case of the Shields parameter, both hrms and hmax esti-
mates are larger when using decomposition (25). However they still remain within the ripple regime. Using
decomposition (1) hrms is only 12% different than the grain roughness Shields parameter (Table 1). Finally,
the friction factor estimates for this mobile sediment bed and prescribed wave conditions is 0.05, regard-
less of the decomposition scheme.

6. Summary and Concluding Remarks


The unsteady nearbed flow field over a highly mobile rippled sediment bed subjected to free-surface grav-
ity waves was examined using the unsteady DANS equations. This technique provided for a stress partition-
ing scheme in which individual stress and momentum transfer terms (viscosity, waves and currents,
turbulence, and bed form-induced) can be independently assessed.
The analysis suggests that (1) viscous stresses are negligible; (2) the total stress near the bed is dominated
by wave and current-induced momentum transfer mechanisms; (3) turbulent and bed form-induced
stresses represent the same amount of momentum transfer near the bed; (4) bed form-induced stresses are
correlated to the presence of coherent motions and can be partitioned from turbulence; and (5) the pres-
ence of bed forms affects the mean momentum flux 1–1.5gb above the ripple crest.
Agreement between the total stress and the bed load transport model of Meyer-Peter and Muller [1948] sug-
gests that the total momentum transfer can generally resolve the ripple mobilization. Additionally, when
appropriately nondimensionalized, stress values are consistent with common sediment transport (h2:5 )
parameterizations. Using the DANS technique and a linear regression to the common quadratic stress
model yields a friction factor of 0.053 6 0.009, slightly larger than f2:5 . Inconsistencies between the total
stress and the local momentum transfer computed from single velocity profiles across the ripple suggest
that an array of spatially distributed sensors is necessary to accurately infer the nearbed dynamics and bed
form evolution.
The computable terms constituting the momentum balance presented in (16) (i.e., the acceleration deficit,
stress gradient, and bed form-induced skin friction) provide an estimate of the unknown bed form-induced
pressure term that is consistent with flow separation. The results suggest the existence of different regions:
(1) the free-stream, where all terms are relatively negligible; (2) the nearbed (1-1.5gb), where an imbalance
of momentum is seen at phases of flow weakening; and (3) below the ripple crest, where stress gradients
are low and bed form-induced pressure dominates.
Decomposition of the large-scale dynamics was explored by further decomposing the large scale terms into
a mean (current) and a periodic (wave) component. RMS stresses and momentum balance terms computed

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8549
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

at the ripple crest elevation are fairly consistent across decomposition schemes, with slightly larger esti-
mates for the decomposition including waves and currents. Estimates of the Shields parameter are also
slightly larger when the formulation decomposes waves and currents. Very little effect on the wave friction
factor was found regarding the choice of decomposition technique.
This effort provides new insights into wave bottom boundary layer dynamics from a physics-based analysis
technique implemented on detailed observations of the nearbed flow field. The technique is based on
double-averaging (in time and space) the flow field near a rough boundary. Our analysis included a time
averaging domain of thirty waves, and a space averaging domain of one ripple wavelength. It is recom-
mended that future efforts expand the observational capabilities to include larger spatial domains.

Acknowledgments References
This work was supported by the
Barr, B. C., D. N. Slinn, T. Pierro, and K. B. Winters (2004), Numerical simulation of turbulent, oscillatory flow over sand ripples, J. Geophys.
National Science Foundation (CTS-
Res., 109, C09009, doi:10.1029/2002JC001709.
0348203) and the National Defense
Carstens, M., F. M. Neilson, and H. D. Altinbilek (1969), Bedforms generated in the laboratory under an oscillatory flow: Analytical and
Science & Engineering Graduate
experimental study, CERC Tech Memo 28, US Army Corps of Engineers, Coastal Engineering Research Centre.
fellowship. The authors are very
Cheng, N. (2009), Comparison of formulas for drag coefficient and settling velocity of spherical particles, Powder Technol., 189(3), 395–398.
grateful for the assistance provided by
members of TUDelft’s Fluids Coleman, S. E., V. I. Nikora, and T. Schlicke (2008), Spatially-averaged oscillatory flow over a rough bed, Acta Geophys., 56(3), 698–733.
Mechanics Laboratory (M. Henriquez, Dick, J., M. Erdman, and D. Hanes (1994), Suspended sand concentration events due to shoaled waves over a flat bed, Mar. Geol., 119(1),
A. Reniers, and M. Stive), OSU’s Coastal 67–73.
Sediment Transport Laboratory (G. Engelund, F., and E. Hansen (1967), A Monograph on Sediment Transport in Alluvial Streams, 62 p., Denmark Tech. Univ., Copenhagen.
Smith, G. Margelowsky, and D. Frank), Fernandez, R., J. Best, and F. L opez (2006), Mean flow, turbulence structure, and bed form superimposition across the ripple-dune transi-
and UNH’s Center for Ocean tion, Water Resour. Res., 42, W05406, doi:10.1029/2005WR004330.
Engineering (J. Irish, T. Lippmann, and Foster, D. L., R. A. Beach, and R. A. Holman (2000), Field observations of the wave bottom boundary layer, J. Geophys. Res., 105(C8), 19,631–
M. Wengrove). The authors particularly 19,647.
appreciate insightful suggestions Foster, D. L., A. J. Bowen, R. A. Holman, and P. Natoo (2006), Field evidence of pressure gradient induced incipient motion, J. Geophys. Res.,
made by S. Coleman regarding the 111, C05004, doi:10.1029/2004JC002863.
bed form-induced stress, and the Fredsøe, J., K. H. Andersen, and B. M. Sumer (1999), Wave plus current over a rippled-covered bed, Coastal Eng., 38, 177–221.
valuable feedback provided by two Gimenez-Curto, L. A., and M. A. Corniero Lera (1996), Oscillating turbulent flow over very rough surfaces, J. Geophys. Res., 101(C9), 20,745–
anonymous reviewers and L. Gimenez- 20,758.
Curto regarding the method of Gimenez-Curto, L. A., and M. A. Corniero (2002), Flow characteristics in the interfacial shear layer between a fluid and a granular bed, J. Geo-
decomposition. The data in this paper phys. Res., 107(C5), 3044, doi:10.1029/2000JC000729.
may be obtained from the Grant, W. D., and O. Madsen (1979), Combined wave and current interaction with a rough bottom, J. Geophys. Res., 84(C4), 1797–1808.
corresponding author (S. Rodriguez- Grant, W. D., and O. S. Madsen (1986), The continental shelf bottom boundary layer, Annu. Rev. Fluid Mech., 18, 265–305.
Abudo, rodriguez.abudo@upr.edu) free Hanes, D., V. Alymov, Y. Chang, and C. Jette (2001), Wave-formed sand ripples at duck, North Carolina, J. Geophys. Res., 106(C10), 22,575–
of charge upon request. 22,592.
Haque, M., and K. Mahmood (1983), Analytical determination of form friction factor, J. Hydraul. Eng., 109, 590–610.
Henriquez, M., A. J. H. M. Reniers, B. G. Ruessink, M. J. F. Stive, T. P. Stanton, and D. L. Foster (2008), On the scaling of sediment transport in
the nearshore, in International Conference on the Application of Physical Modelling to Port and Coastal Protection, edited by L. Damiani
and M. Mossa, pp. 1–8, Int. Assoc. of Hydro-Environ. Eng. and Res., Bari, Italy.
Hino, M., M. Kashiwayanagi, A. Nakayama, and T. Hara (1983), Experiments on the turbulence statistics and the structure of a reciprocating
oscillatory flow, J. Fluid Mech., 131, 363–400.
Hsu, T., T. Sakakiyama, and P. Liu (2002), A numerical model for wave motions and turbulence flows in front of a composite breakwater,
Coastal Eng., 46(1), 25–50, doi:10.1016/S0378-3839(02)00045-5.
Hsu, T.-J., and D. M. Hanes (2004), Effects of wave shape on sheet flow sediment transport, J. Geophys. Res., 109, C05025, doi:10.1029/
2003JC002075.
Jonsson, I. (1966), Wave boundary layers and friction factors, in Proceedings of the 10th International Conference on Coastal Engineering, pp.
127–148, Univ. of Tokyo, Tokyo.
Lofquist, K. E. (1980), Measurements of oscillatory drag on sand ripples, Coastal Eng. Proc., 1(17), 3087–3108.
Lowe, R., J. Koseff, and S. Monismith (2005), Oscillatory flow through submerged canopies: 1. Velocity structure, J. Geophys. Res., 110,
C10016, doi:10.1029/2004JC002788.
Maddux, T., S. McLean, and J. Nelson (2003), Turbulent flow over three-dimensional dunes: 2. Fluid and bed stresses, J. Geophys. Res.,
108(F1), 6010, doi:10.1029/2003JF000018.
Maier, I., and A. Hay (2009), Occurrence and orientation of an orbital ripples in near-shore sands, J. Geophys. Res., 114, F04022, doi:10.1029/
2008/JF001126.
Meyer-Peter, E., and R. Muller (1948), Formulas for bed-load transport, in Report on the Second Meeting of the International Association for
Hydraulic Structures Research, pp. 39–64, Int. Assoc. of Hydro- Environ. Eng. and Res., Stockholm.
Nichols, C. S., and D. L. Foster (2007), Full-scale observations of wave-induced vortex generation over a rippled bed, J. Geophys. Res., 112,
C10015, doi:10.1029/2006JC003841.
Nichols, C. S., and D. L. Foster (2009), Observations of bed form evolution with field-scale oscillatory hydrodynamic forcing, J. Geophys. Res.,
114, C08010, doi:10.1029/2008JC004733.
Nielsen, P. (1992), Coastal Bottom Boundary Layers and Sediment Transport, World Sci., Singapore.
Nikora, V., D. Goring, I. McEwan, and G. Griffiths (2001), Spatially averaged open-channel flow over rough bed, J. Hydraul. Eng., 127,
123–133.
Nimmo Smith, W. A. M., J. Katz, and T. R. Osborn (2005), On the structure of turbulence in the bottom boundary layer of the coastal ocean,
J. Phys. Oceanogr., 35, 72–92.
Penko, A., J. Calantoni, S. Rodriguez-Abudo, D. Foster, and D. Slinn (2013), Three-dimensional mixture simulations of flow over dynamic rip-
pled beds, J. Geophys. Res. Oceans, 118, 1543–1555, doi:10.1002/jgrc.20120.

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8550
Journal of Geophysical Research: Oceans 10.1002/2014JC010240

Raffel, M., C. Willert, S. Wereley, and K. Jurgen (2007), Particle Image Velocimetry: A Practical Guide, Springer, Berlin.
Raudkivi, A. (1966), Bed forms in alluvial channels, J. Fluid Mech., 26(3), 507–514.
Reynolds, W., and A. Hussain (1972), The mechanics of an organized wave in turbulent shear flow. part 3. Theoretical models and compari-
sons with experiments, J. Fluid Mech., 54(02), 263–288.
Rodriguez-Abudo, S., D. L. Foster, and M. Henriquez (2013), Spatial variability of the wave bottom boundary layer over movable rippled
beds, J. Geophys. Res. Oceans, 118, 3490–3506, doi:10.1002/jgrc.20256.
Shen, W. H., H. M. Fehlman, and C. Mendoza (1990), Bed form resistances in open channel flows, J. Hydraul. Eng., 116(6), 799–815.
Sleath, J. F. A. (1987), Turbulent oscillatory flow over rough beds, J. Fluid Mech., 182, 369–409.
Sleath, J. F. A. (1999), Conditions for plug formation in oscillatory flow, Cont. Shelf Res., 19, 1643–1664.
Smith, J. D., and S. R. McLean (1977), Spatially averaged flow over a wavy surface, J. Geophys. Res., 82(12), 1735–1746.
Sveen, J. K. (2004), An Introduction to MatPIV 1.6.1, 2nd ed., Dep. of Math., Univ. of Oslo, Oslo.
Swart, D. H. (1974), Offshore sediment transport and equilibrium beach profiles, Delft Hydraul. Lab. Publ. 131, Delft.
van der Dominic, A., T. O’Donoghue, A. Davies, and J. Ribberink (2011), Experimental study of the turbulent boundary layer in acceleration-
skewed oscillatory flow, J. Fluid Mech., 684, 251–283.
Van der Werf, J. J., J. S. Doucette, T. O’Donoghue, and J. S. Ribberink (2007), Detailed measurements of velocities and suspended sand con-
centrations over full-scale ripples in regular oscillatory flow, J. Geophys. Res., 112, F02012, doi:10.1029/2006JF000614.
Wilson, N., and R. Shaw (1977), Higher-order closure model for canopy flow, J. Appl. Meteorol., 16(11), 1197–1205.
Zhou, J., R. J. Adrian, S. Balachandar, and T. M. Kendall (1999), Mechanisms for generating coherent packets of hairpin vortices in channel
flow, J. Fluid Mech., 387, 353–396.

RODRIGUEZ-ABUDO AND FOSTER C 2014. American Geophysical Union. All Rights Reserved.
V 8551

You might also like